Post on 06-Aug-2020
Predictions, Data and ICH M7
Sales Manager
Krishan Patel
Krishan.Patel@lhasalimited.org
In silico workflow under M7
Expert Review
2 in silico predictionsexpert + statistical
Databases, in-house, literature..
Known mutagen
Both predict positive
Both predict negative
Ames testLimit according to TTC or present purge argument for loss
Treat as non-mutagenic
Knownnon-mutagen
Disagree / fail to predict
Evaluate drug substance, impurities, degradants, intermediates…
Agenda
• Derek Nexus and Sarah Nexus• Introduction
• Demonstration
• Differences between Derek Nexus and Sarah Nexus
• Lhasa’s ICH M7 Tool• Demonstration
• Questions
What is Derek?
• Knowledge based expert system
• Structure Activity Relationship (SAR)
• Rules written by scientists from detailed expert analysis
• Two step process:
✓ Identifies toxicophores
✓ Makes qualitative predictions
• Transparent – provides supporting references &
examples
Alerts
• An Alert is a set of structural features in a molecule,
that make a user suspect that the substance may show
a particular effect
• Currently, the Certified Derek Knowledge Base has 890
alerts and 74 endpoints with 9 parent endpoints
• 134 of these alerts are for mutagenicity
Knowledge base search
Skin permeability
Presence oftoxicophore
Molecular weight
Species
Metabolicactivation
Demonstration
What is Sarah?
Statistical approach for prediction of mutagenicity
Positive and negative predictions
Defined Applicability Domain
Transparent – supported with example structures
Methodology
Fragment
Dictionary
Fragmentation
Decision tree
leading to
Hypotheses
Hypothesis
Mining
Self
OrganisationHierarchical Network
SOHN
Ames Experimental
Data (curated)
Demonstration
Differences Between Derek and Sarah
Data
• Uses all Lhasa data
• Includes consortia &
donated confidential data +
data mined on-site
• Only uses non-confidential
data
Methodology
• Expert system
• Human-written rules based
upon data & knowledge
• Statistical model
• Machine-learning model using
a hierarchical network
Scope of alert • Hand-written Markush • Fragments learnt by model
Interpretability
• References
• Expert commentary
• Mechanistic explanation
• Some supporting examples
• Transparent methodology
• Learning summarised by
hypothesis
• Direct access to training set
• Confidence in prediction
M7 Functionality
• With a Derek and Sarah licence you can run an ‘ICH M7
prediction’
• Allows you to quickly see if the two systems agree or
disagree
• One Click Reporting
Demonstration
Any questions?